OBIEE Deployment & Change Management Mark Rittman, Technical Director, Rittman Mead Rocky Mountains Oracle User Group Training Days 2012, Denver
Mark Rittman Mark Rittman, Co-Founder of Rittman Mead Oracle ACE Director, specialising in Oracle BI&DW 14 Years Experience with Oracle Technology Regular columnist for Oracle Magazine Author of forthcoming Oracle Press book on OBIEE 11g Writer for Rittman Mead Blog : http://www.rittmanmead.com/blog Email : mark.rittman@rittmanmead.com Twitter : @markrittman
About Rittman Mead Oracle BI and DW platinum partner World leading specialist partner for technical excellence, solutions delivery and innovation in Oracle BI Approximately 30 consultants worldwide All expert in Oracle BI and DW UK based Offices in US, Europe (Belgium) and India Skills in broad range of supporting Oracle tools: OBIEE OBIA ODIEE Essbase, Oracle OLAP GoldenGate Exadata
Oracle Business Intelligence 11g (11.1.1.5) Oracle s BI platform, now at release 11.1.1.5 (11gR1) Wide range of servers, tools, metadata stores based around Oracle FMW11g Based on Siebel Analytics with additions from Oracle and Hyperion products Often used in conjunction with the BI Applications and EPM Suite
Elements of an Oracle BI EE 11g Project Oracle BI Repository (RPD file) Oracle BI Presentation Catalog System Configuration Settings UI Customizations Security artifacts (application roles, users, directory settings) Plus associated database schemas, ETL packages etc (out of scope for this though)
OBIEE 11g Project Lifecycle Stage #1 - Early Days Prototype to first production Typically a single developer, no version-control Initial project is moved from DEV server into PROD once first phase complete Single Developer End Users Full upload of RPD and catalog via EM Full upload of RPD and catalog via EM Dev Test Prod
OBIEE 11g Project Lifecycle Stage #2 - Further Releases Updates to this first release, to add new RPD objects, shared folder catalog objects Incremental metadata needs to be merged into PROD, keeping existing objects Uses the three-way merge features for the RPD and the catalog Single Developer End Users Full upload of RPD and catalog via EM Incremental update of RPD and catalog shared folders via merging, then EM M e r g e Dev Test Prod
OBIEE 11g Project Lifecycle Stage #3 - Project Expands Out Additional developers wish to add content to the RPD Typically all developers on the project start accessing the RPD online, concurrently Other separately developed projects may need to be merged into the main RPD Version control becomes important as multiple developers start contributing changes Source Control Multiple Developers End Users Other project RPDs M e r g e M e r g e Dev Test Prod
OBIEE 11g Project Lifecycle Stage #4 - Enterprise Deployment Soon, ad-hoc merging of RPDs and shared online development becomes unworkable A system needs to be put in place to handle distributed development Multi-User Development (MUD) Environment then becomes an option Source Control Multiple Developers End Users M e r g e Source Control Development Branches Source Control M e r g e MUD Administrator Test M e r g e Prod
Propagating System Configuration Changes At various points, system configuration changes have to be applied to BIEE environments Deploying new repositories, presentation catalogs Enabling SSL, new connections to directories (AD) etc Changing performance parameters All changes have to be applied to all nodes in a cluster, possibly with rolling-restarts Developer Production Support Dev Test Production Cluster
BI EE Features to Support Change Management & Deployment Three-way merges of repository files Catalog archiving/unarchiving New in 11g - repository and catalog patching Multi-User Development Environment New in 11g - Enterprise Manager Fusion Middleware Control ( EM ) New in 11g - WebLogic Server Scripting Tool & Oracle BI Systems Management API
Merging Repository Files (RPDs) Merging repositories is a common task on projects past the initial stage To merge new and changed objects in DEV into the PROD repository To merge two RPDs into one, to run online in PROD Common task in general software development projects, with common complications Repositories may contain similarly-named objects, but logically different Repository objects may have changed in both DEV and PROD - which do you choose? These, and others, are called merge conflicts Repository #1 Repository #2 Merged Repository
Oracle BI Repository Three-Way Merges Oracle BI, like many software development tools, uses the concept of three-way merges A modified repository, which is usually the PROD repository A current repository, which is usually the DEV repository An original repository, from which they were both derived Provides a number of benefits compared to 2-way merges Avoids guessing whether objects with the same name are actually logically the same For branching development, allows both branches to be updated with subsequent changes to the original More accurate and efficient way of merging two sets of objects with common parentage If no common repository available, then substitute blank repository (and loose the 3-way merge benefits) Repository #1 ( current ) Common Parent Repository ( original ) Merged Repository Repository #2 ( modified )
Understanding Merge Rules The merge process makes most sense when you understand the merge rules The RPDs you select for modified and current are important, do not choose at random Current = development, Modified = production Rules assume that changes added to modified want to be preserved Deletions in current that are still in modified have to be confirmed Additions added, or deletions from, both repositories are automatically propagated Objects added to both, but with differences, cause a merge conflict Objects modified in both cause a merge conflict Common Parent Repository ( original ) Repository #2 ( current ) Development M e r g e Merged Repository New Production Repository #1 ( modified ) Production
Repository Equalization Repositories may contain logically identical objects that have different upgrade IDs Upgrade IDs are internal ID codes for objects in the repository Can be caused by deleting, then recreating, the same object Mismatching upgrade IDs can cause the upgrade process to create duplicates in the merge repository, thinking that the two objects are completely different Answer is to equalize the repositories Modified Repository (Production) Sales Subject Area Upgrade ID : 1001 Current Repository (Development) Sales Subject Area Upgrade ID : 1021 E q u a l i z a t i o n Modified Repository (Production) Sales Subject Area Upgrade ID : 1001 Current Repository (Development) Sales Subject Area Upgrade ID : 1001 M e r g e Merged Repository Sales Subject Area Upgrade ID : 1001
Three-Way RPD Merge Step 1 : Open Current RPD, Select Merge Start the BI Administration tool Open the Current (typically, Development) repository offline Select File > Merge...
Three-Way RPD Merge Step 2 : Select Modified & Current RPDs With the Merge Repository Wizard - Select Import Files dialog open, select the modified (production) and original (common parent) RPDs Enter passwords Tick the Equalize during merge checkbox Select Full Repository Merge
Three-Way RPD Merge Step 3 : Resolve Conflicts Conflicts typically occur if a choice needs to be made between two options Select the choice either from the modified (prod) or current (dev) repository Choices can go down to the object property level Once all conflicts resolved, merged repository is then opened for editing
New in 11g : Repository and Catalog Patching In some situations, you want to perform the merge hands-off To remove opportunity for human error, to allow it to be scripted 11g introduces the concept of repository (RPD) and catalog patching Whole process, from extracting changes to patching target, can be scripted Development Repository Production Repository Development Catalog Production Catalog
Repository Patching Repository patching is a two-step process 1. Compare the current repository to the original one, create an XML patch file of the differences between the two 2. Apply the XML patch file to the modified repository, as a three-way merge also using the original repository Can be performed using BI Administrator tool, or from the command-line 1 2 Repository #2 ( current ) Development Common Parent Repository ( original ) C o m p a r e XML Patch File of diffs between Current and Original RPD Repository #1 ( modified ) Production Common Parent Repository ( original ) M e r g e Merged Repository New Production XML Patch File
Interactive Repository Patching using the BI Administration Tool To create the XML patch file, use the File > Compare... feature To apply the patch file, use the File > Merge... feature
Patching using BI Administrator Step 1 : Open Current RPD Usign the BI Administrator tool, open the current (development) repository offline Select File > Compare...
Patching using BI Administrator Step 2 : Open Current RPD View comparison report Press Equalize button if required
Patching using BI Administrator Step 3 : Generate Patch File Press the Create Patch... button to create the XML patch file
Patching using BI Administrator Step 4 : Open Modified RPD Now open the modified (production) repository, select File > Merge
Patching using BI Administrator Step 5 : Select Patch and Original With the Merge Repository Wizard - Select Import Files dialog open, select the original (common parent) RPD and the XML patch file Patch file substitutes for the current repository Select Patch Repository
Patching using BI Administrator Step 6 : Resolve any Conflicts As with full repository merges, you may then need to resolve merge conflicts Same rules around current, modified and original RPD merge rules apply Once resolved, merged repository is then opened for editing
Command-Line Creation and Applying of RPD Patches Command-line utilities are avaialble for creating, and applying, RPD patch files comparerpd creates a patch file based on current and original repositories comparerpd P [current repository password] C [current repository path and name] W [original repository password] G [original repository path and name] D [patch file path and name] patchrpd does a three-way merge with the patch file, original and modified repositories patchrpd -P [modified repository password] -C [modified repository path and name] -Q [original repository password] -G [original repository path and name] -I [patch file path and name] -O [new repository path and name] Both located at [middleware_home]\oracle_bi1\bifoundation\server\bin\
Version Control in Software Development Projects Version control is a common concept in software development Allows you to store copies (versions) of project elements over time Refer back to old versions, restore old versions, create named/numbered releases Branch projects, re-combine branches Typically peformed using tools such as PVCS, Subversion, Git, Visual Sourcesafe Version Control System RPD # 1.22 RPD # 1.30 RPD # 1.35 RPD # 2.45 RPD # 2.47 Download working copy Check-in changes Retrieve historical versions at will Local copy of RPD Version # 2.47
Subversion and OBIEE 11g OBIEE 11.1.1.5 does not have in-built integration with version or source control But you can store the various project artifacts in any version control tool Subversion, together with VisualSVN Server and TortoiseSVN, are suitable tools There are however some limitations The RPD has to be uploaded in its entirety Although you can also upload XML patch files The Catalog has to be archived before uploading You cannot use the merge/patch facility in SVN, you must use BI Administrator / Catalog Manager patch/merge instead
Creating a Subversion Repository for use with Oracle BI EE A standard subversion repository should be created for the OBIEE project Create standard three directories Trunk : the main development path for the project Branches : branches off of the main development path Tags : named/numbered releases All top-level directories have subdirectories for RPD, catalog, config etc Trunk RPD Catalog Project Name Branches Config Repository Name Project Name Tags Project Name
SVN Development Lifecycle Step 1 : Checkout of Project Files Developer right-clicks on desktop, selects SVN Checkout... Select trunk folder, or particular branch or tag Select HEAD revision, or particular revision number Files copied to version-controlled folders on desktop 1 2 3
SVN Development Lifecycle Step 2 : Edit Files, Upload to EM Files are then usually copied to a working directory, or uploaded to EM for online edits Perform all changes as required, add or delete objects Working copies can be updated using SVN Update Objects are not specifically locked when you check-out, unless you choose to lock them Ensure all changes are performed in relevant tools (BI Admin, Catalog Manager etc) Once changes complete, copy back to SVN folders on desktop Ensure catalog is archived, not left as folder + files
SVN Development Lifecycle Step 3 : Check Changes Back into SVN Once complete, new and changed files can be added back into SVN repository Right-click on folders, select SVN Commit... New files have to be added to the project to be included in check-in Uploaded files get a new revision number Text files are just stored as diffs Binary files (RPD, catalogs etc) are stored in their entirety 1 2 3
Tagging Projects Particular revisions/versions can be tagged as a particular release Version 1.0 Version 2.0 migrated to OBIEE 11.1.1.5 Check out the top-level folders for a project (e.g. GCBC_OBIEE) for a particular revision Right-click on the Trunk folder and select TortoiseSVN > Branch/Tag Copy the folder to a new sub-folder under the Tags directory (for example, Rel. 1.0) 1 2 3
Branching Projects Projects can be branched in the same way as tagged Create new sub-folder under the Branches folder Used for when a copy of the project is made, then enhanced separately For rolling out country-specific versions, with localizations For working through an upgrade to OBIEE 11gR2, whilst still preserving the 11gR1 version If you wish to merge the branch back into the Trunk folder, use BI Admin three-way merge rather than SVN s merge/patch feature UK-Localized Branch Main Development Stream Repository Optimization Branch Test Upgrade Branch
Managing Change with Large Teams of Developers So far, we have looked at projects where there is a single developer On many projects though, you wish to scale-up developers to deliver larger scope The catalog supports multiple developers editing, adding objects etc For smaller teams, you might consider concurrent online editing of the RPD Has the virtue of simplicity 11g certifies up to 5 concurrent developers Works through a system of check-out and check-in of objects - Check-out is coarse-grained though - Edits to a logical table lock the whole business model Multiple Developers End Users Source Control Dev Test Prod
Multi-User Development Environment (MUD) MUD Administrator divides main repository into projects; self-contained RPD subsets Master repository is then published to a network share Projects are then worked on independently, and then merged back into the master RPD Uses the repository compare and merge features under the covers Works best when each developer has a full OBIEE Sandbox environment to develop with and unit test their work License considerations through - may require named user plus licensing to be financially viable More complex than online development, but makes sense when you know how it works
What Happens During MUD Check-Out / Merge / Check-In? Developer Selects Projects(s) for checkout from Master RPD Subset RPD + copy of Subset RPD copied to developer PC Developer then edits the RPD, adds, makes changes Subset RPD is then compared to the subset RPD copy Local changes merged in to local copy of the master RPD, merge conflicts resolved Changes are published to the network Master RPD, and lock taken during this merge
What Happens During MUD Check-Out / Merge / Check-In? Developer Selects Projects(s) for checkout from Master RPD Subset RPD + copy of Subset RPD copied to developer PC Developer then edits the RPD, adds, makes changes Subset RPD is then compared to the subset RPD copy Change in 11g compared to 10g : Locking only now takes place at the publish step, not merge of local changes Local changes merged in to local copy of the master RPD, merge conflicts resolved Changes are published to the network Master RPD, and lock taken during this merge
Setting up MUD Environment Step 1 : Define Projects Administrator opens the repository to be shared, offline Select Manage > Projects Define projects using selections of business model, or subject area, fact tables Add init blocks, variables, users and other objects Physical layer objects get added automatically
Setting up MUD Environment Step 2 : Copy to Network Share Create network shared directory that is accessible to all developers Set permissions so other users can write to it Copy the file to this directory; this is now the master repository
Setting up MUD Environment Step 3 : Configure Workstations On each developer workstation, configure it for MUD access Create a mapped network drive to the MUD network share folder Select Tools > Options > Multiuser Use Browse... to select the mapped network drive Type in the name of the developer (for recording MUD updates, checkouts, locks)
MUD Lifecycle Step 1 : Select and Checkout Project From the developer workstation, select File > Multiuser > Checkout... Select the project(s) to check-out Name the subset RPD file (a temporary copy of the whole RPD is made here) On save the subset RPD, plus a duplicate, is saved and the temporary copy removed 1 3 2 4
MUD Lifecycle Step 2 : Make Changes to Subset, Do Compare Make changes to the subset RPD, such as adding, deleting or modifying objects After a time, use File > Multiuser > Compare with Original... Performs an automatic File > Compare... with the duplicate subset RPD Save your changes as normal Upload to a sandbox OBIEE environment and run online if required
MUD Lifecycle Step 3 : Merge in Local Changes Once work is complete, select File > Compare > Merge in Local Changes Merges your subset RPD in with a fresh copy of the master RPD Peformed using an automatic three-way merge If there are merge conflicts, this is where you will deal with them Merged results are then stored locally until published by the next step Change compared to 10g: lock is not taken at this step Define Merge Strategy only displayed if merge conflicts encountered
MUD Lifecycle Step 4 : Publish or Discard Changes After local changes have been merged into local copy of the master repository, these changes can then be merged in with the actual master repository Again performed using an automatic three-way merge Changes can also be discarded, or rolled-back (giving you the original subset RPD again) At this point, the lock is taken (to stop multiple sessions trying to write to the master) Only taken briefly in 11g as in most cases, conflicts dealt with in previous step If master RPD has been updated since local merge, local merge is rolled-back and performed again
MUD Lifecycle Step 5 : Viewing MUD History Developers with MUD configured on their workstations can view the MUD history See history of checkouts, check-ins, comments added during publish (lock) phase Useful history of MUD activity
MUD and Version Control If MUD subset RPDs are part of the same development stream, store them as subdirectories under the main RPD directory If MUD is used for branching the project, create a branch using SVN, and then update the branch s RPD with the one generated by the MUD branch check-out RPD MUD Subset RPD #1 Trunk Catalog MUD Subset RPD #2 Project Name Branches Config MUD Subset RPD #3 Repository Name Project Name Tags Project Name
Deploying Configuration Changes across Clustered OBIEE Systems Another aspect of managing change and deployments is at a system level How do you apply configuration changes across multiple clustered nodes? How do you deploy repositories when your BI servers are clustered? How do you script the process so that it is automated? Developer Production Support Dev Test Production Cluster
Project Deployment and Migration Best Practices 1. Use Enterprise Manager to deploy repositories and catalogs between environments 2. Use Enterprise Manager to apply system configuration changes to environments 3. Use WLST and the Oracle BI Systems Management API to script these tasks cd (biinstance.tostring()) biserver = get('serverconfiguration') cd('..') cd(biserver.tostring()) ls() argtypes = jarray.array(['java.lang.string', 'java.lang.string'],java.lang.string) argvalues = jarray.array(['c:/sampleapplite.rpd', 'Admin123'],java.lang.Object) invoke('uploadrepository',argvalues,argtypes) cd('..') cd('oracle.biee.admin:type=bidomain,group=service') objs = jarray.array([],java.lang.object) strs = jarray.array([],java.lang.string) invoke('commit',objs,strs)
Managing the Oracle BI Repository and Web Catalog using EM Enterprise Manager is now used to deploy new RPD files (repository) and presentation catalog directories RPD files are uploaded using EM; catalogs have to be manually copied to servers Deploys metadata across all BI Server and Presentation Server nodes in the cluster (unless shared directories have been defined)
Managing the Oracle BI Repository and Web Catalog using EM Enterprise Manager is now used to deploy new RPD files (repository) and presentation catalog directories RPD files are uploaded using EM; catalogs have to be manually copied to servers Deploys metadata across all BI Server and Presentation Server nodes in the cluster (unless shared directories have been defined)
System Configuration Changes using Enterprise Manager Most important system configuration settings are now managed through EM Ensures that all changes you make are applied across all nodes in the cluster Graphical interface for managing common settings including Caching and other performance settings Number and scale-out of system components across cluster Miscelaneous settings including # rows returned, read-only RPD etc Each BI environment has its own EM website, which manages all nodes in the domain
Manually Managed System Configuration Settings Settings not managed by EM have to be manually managed by editing configuration files (NQSConfig.INI, instanceconfig.xml) etc Populate Aggregate Rollup Hits (Cache) Use Advanced Hit Detection (Cache) Maximum Subexpression Search Depth (Cache) Use Advanced Hit Detection (Cache) Virtual Table Page Size (for in-memory joins, calcs) Be sure to deploy changes across all system components on all nodes, and to not alter managed settings
Summary Projects that scale beyond a single developer need deployment & change management Many tools are available within OBIEE 11g to handle multi-developer teams Keep things as simple as possible; but if required, there is MUD Key to MUD is understanding what goes on when you check-out/check-in projects 11g introduces far less intrusive locking, makes MUD more viable The lack of in-built version control can be overcome with tools such as Subversion Always use EM to propagate system changes, and if required, script with WLST.
More Information Thank you for attending this presentation More information can be found at http://www.rittmanmead.com Contact us at info@rittmanmead.com or mark.rittman@rittmanmead.com Look out for our book, Oracle Business Intelligence Developers Guide due Q1 2012 Follow-us on Twitter (@rittmanmead) or Facebook (facebook.com/rittmanmead)
OBIEE Deployment & Change Mgmt Best Practices Mark Rittman, Technical Director, Rittman Mead Rocky Mountains Oracle User Group Training Days 2012, Denver